Decentralization versus Centralization: What Ensures Food Security? Empirical Evidence from 170 Prefecture-Level Cities in China’s Major Grain-Producing Areas

Abstract

Whether fiscal decentralization will lead to agricultural land “non-grainization” has been widely debated in academic circles. How to improve the efficiency of financial support to agriculture and optimize the grain planting structure by clarifying the relationship between central and local powers and responsibilities is the key to ensuring food security. Based on the panel data of 170 cities in China from 2004 to 2017, this paper uses system moment estimation and a threshold effect model to explore the impact of fiscal decentralization on grain planting structure. The results show that (1) fiscal decentralization has a significant negative effect on the share of food crop cultivation in the major grain-producing areas. (2) Taking the wage level, financial support for agriculture, and land finance as the threshold variables, the test finds that there is a threshold effect of fiscal decentralization on the proportion of food crop cultivation, in which land finance dependence and rises in the wage level are conducive to mitigating the negative effect of fiscal decentralization on the proportion of food crop cultivation. (3) For the three major types of food crop varieties, the negative impact of fiscal decentralization on the share of wheat and corn crop cultivation is subject to the threshold effect of wage level, financial support for agriculture, and land finance, while the impact of fiscal decentralization on the share of rice crop cultivation is not significant. The results of the study have an important guiding role for the government to deepen the reform of the tax-sharing system, improve the long-term mechanism of stable growth of financial support for grain, and optimize the layout of the grain industry.

1. Introduction

Food security is crucial for global stability, human well-being, and sustainable development. The latest “State of Food Security and Nutrition in the World” report released by the Food and Agriculture Organization of the United Nations highlights a severe global famine situation, with malnutrition worsening significantly. The total number of people facing hunger has surpassed 735 million, a sharp increase from 613 million in 2019, with the proportion of the population at food crisis levels rising from 25.3% in 2019 to 29.6% in 2022. Following the outbreak of COVID-19, the number of countries at risk of food supply disruptions escalated from 53 pre-pandemic to 79 in 2023, putting as many as 783 million people at risk of long-term hunger. The World Food Programme’s assessments in 78 countries suggest that in 2023, more than 333 million people were facing severe food crises globally, an increase of nearly 200 million compared to the pre-pandemic levels.

As an essential for national survival and health, food production inevitably requires government fiscal regulation and support. In China, fiscal decentralization, as an institutional arrangement of central–local power distribution, significantly determines the efficiency and method of local government fiscal resource allocation. The 1994 tax-sharing reform in China effectively addressed the financial and administrative relations between central and local governments. Concurrently, as the constraints on resources and environment for food production tightened, China initiated a new era of food security strategies. Local governments, under the provincial governor responsibility system for food security and central special transfer payments, have increasingly strengthened fiscal support for agriculture. Therefore, examining the link between fiscal decentralization and agricultural planting structure has significant theoretical and practical relevance for refining China’s fiscal decentralization and achieving new-era food security strategies.

Since the central government’s No. 1 document in 2017 proposed agricultural supply-side structural reforms, more studies have begun focusing on structural reforms in the planting industry. Previous research on grain planting structure has been conducted at macro, micro, and policy levels. For instance, studies have analyzed the spatial dynamic changes in grain structure, finding significant shifts in the centers of increased production for rice, wheat, corn, and soybeans [1,2,3], with soybeans transitioning to corn, while rice remains stable and the grain planting concentration grows [4,5]. This research offers a macro-perspective on the spatial distribution and evolving trends in grain planting structures, laying a foundation for further investigations into the causes and impacts of these changes.

Other studies have examined the impact of food subsidy policies on grain planting structures, indicating that moderate subsidy policies can directly affect grain planting structures and scales [6,7,8,9]. For example, scholars have noted the positive role of China’s “four subsidies” policy in promoting grain production [10] and have assessed the performance of agricultural subsidy policies, highlighting variations in policy effectiveness [11] and the resource allocation effects of price support policies on soybean total factor productivity [12].

Research has also explored the factors influencing farmers’ decisions towards “grainization” or “non-grainization” of planting structures, including income structure [13,14], human capital [15,16], and comparative planting profits [17,18]. These studies reveal how individual farmer behaviors, influenced by personal characteristics, market conditions, and policies, affect grain planting structures, offering insights into optimizing grain planting structures by altering farmer behaviors.

Furthermore, the relationship between fiscal decentralization and grain planting structures centers on the rational division of fiscal revenue, government functions, and expenditures among different government levels. It has been proven that grain production is closely linked to government fiscal support, significantly determining the effectiveness of grain policies. For instance, the lack of enthusiasm in government support for grain planting is primarily due to unclear responsibilities [19]. Fiscal decentralization leads to excessive financial pressure on local governments, which affects the scale of grain planting [20]. While some scholars argue that fiscal decentralization distorts local government spending structures, reducing the growth rate of agricultural support expenditures, the overall scale of fiscal support for agriculture remains on an upward trend under the dual effects of special transfer payments and the “rice bag governor responsibility system” [21].

As a system, fiscal decentralization can adapt to the specific needs and environments of different countries worldwide. By allowing local governments to formulate economic policies based on local conditions, it facilitates policy diversification and variety, better meeting citizens’ needs. Moreover, fiscal decentralization, by granting local governments greater fiscal authority and autonomy, motivates them to provide public services and promote economic development more actively. On one hand, fiscal decentralization can advance the democratization process, enhancing local governments’ autonomy and sense of responsibility, thereby fostering a focus on public welfare issues and promoting democratization. On the other hand, it can increase government efficiency. Fiscal decentralization enables local governments to be closer to citizens and more knowledgeable about local conditions and needs, thus effectively providing public services and solving local problems. Additionally, fiscal decentralization optimizes resource allocation, allowing local governments to allocate resources based on local realities and demands, enhancing resource use efficiency. This helps alleviate issues of uneven resource distribution and waste, promoting equitable and sustainable human societal development. Thus, exploring the grain planting structure in China’s major grain-producing areas from the perspective of fiscal decentralization holds significant practical importance.

Whether fiscal decentralization will lead to agricultural land “non-grainization” has been widely debated in academic circles. How to improve the efficiency of financial support to agriculture and optimize the grain planting structure by clarifying the relationship between central and local powers and responsibilities is the key to ensuring food security. Therefore, this paper reviews previous studies on fiscal decentralization and grain planting, proposes theoretical hypotheses about the impact of fiscal decentralization on grain planting structures, and validates these hypotheses using panel data from 170 prefecture-level cities in China’s major grain-producing areas from 2004 to 2017, employing system GMM and threshold models. Compared to previous studies, this paper’s marginal contributions are threefold: First, it addresses the gaps in research on fiscal decentralization in agricultural planting structures. Second, focusing on China’s major grain-producing areas, the primary drivers of effective grain supply, gives the study practical significance and provides a reference for other regions. Third, by constructing a threshold effect model using wage levels, agricultural support structure, and land price as threshold variables, it further analyzes the impact of fiscal decentralization on grain planting areas.

2. Theoretical Analysis and Hypotheses

2.1. Fiscal Decentralization and Agricultural Planting Structure

Fiscal decentralization is a pivotal institution for managing the relationship between central and local governments in China and is a key basis for delineating fiscal authority between them. It largely dictates the preferences and efficiency of fiscal resource allocation. However, compared to the clarity of fiscal authority division, the division of governmental responsibilities and expenditure accountability remains ambiguous, leading to a disjunction between authority and responsibility in China’s administrative divisions [22].

In practice, grain production heavily relies on government support; therefore, clarity in governmental responsibilities directly affects grain subsidies and the development of agricultural scale. Currently, the mismatch in food security responsibilities and unclear accountability between central and local governments in China are inadequate to meet the needs of food security under new circumstances [23]. Specifically, first, under the mismatch between fiscal authority and responsibilities, local governments tend to implement land finance policies, driving up land prices, and consequently, reducing arable land area [24], which affects grain production. Second, as governmental responsibilities increase and fiscal authority decreases, the fiscal deficit of local governments expands. Local governments’ spending has an urban bias, aiming to increase fiscal revenue and economic growth, which is detrimental to grain production and its scalable operation [25]. Third, in an assessment system centered on economic performance, local governments with greater fiscal autonomy use their political and economic resources to promote rapid regional economic growth in hopes of achieving political promotion, leading to a neglect of agricultural development and infrastructure construction. This causes a deficiency and misalignment in the supply of quasi-public agricultural infrastructure [26], lowering farmers’ enthusiasm for grain farming. Furthermore, since grain possesses the characteristics of a quasi-public good, it is prone to market failures and free-riding phenomena. For tax revenue, local governments have incentives to encroach on or even misappropriate funds designated for agriculture [27], leading to poor targeting of specialized funds, ultimately hindering grain production.

Based on the above analysis, this paper proposes the following hypothesis:

Hypothesis 1:

Fiscal decentralization will lead to the “non-grainization” of the planting structure.

2.2. The Nonlinear Relationship of Fiscal Decentralization on Agricultural Planting Structure

Building on the theoretical analysis of the impact of fiscal decentralization on grain planting structure, it becomes evident that the influence of fiscal decentralization largely stems from land finance and the intensity of fiscal support for agriculture. The dependency on land finance and the strength of fiscal agricultural support might create variations in the direction or magnitude of fiscal decentralization’s impact on grain planting. Additionally, rising urban wage levels may promote a “voting with their feet” effect among local governments and alter the income structure of farming households, potentially mitigating the “non-grainization” effect of fiscal decentralization on agricultural planting structure. Therefore, this study selects wage levels, fiscal support for agriculture, and land finance as threshold variables to examine the differences in the impact of fiscal decentralization on the area planted with grains across various threshold ranges.

Under a fiscal decentralization regime, local governments possess relatively independent fiscal and administrative authority, affording them significant space for resource allocation. This flexibility allows them to more effectively respond to changes in the grain market, promoting the development of grain production. However, as wage levels increase and the comparative advantage of agriculture declines, this might lead to an increase in the “non-grainization” effect of fiscal decentralization. With increasing urban–rural population mobility, higher urban wage levels can attract rural labor to non-agricultural sectors, enhancing the effectiveness of the “voting with their feet” mechanism [28], thereby impacting the local grain planting structure. On one hand, rising wage levels lead to an increase in non-agricultural labor, and farmers may opt for grain crops that are more amenable to mechanization, reducing the area planted with cash crops [13]. Thus, rising wages and labor shifts can promote the application of agricultural mechanization, which further accelerates the reduction in labor-intensive crops and the increase in less labor-intensive crops [29], thereby increasing the area sown with grains. On the other hand, the shift in agricultural labor to wage-earning employment can promote the development of agricultural productive service organizations, and the development of the agricultural productive service outsourcing market can enhance farmers’ enthusiasm for grain farming, strengthening their tendency to plant grains [30]. Therefore, higher wage levels can promote a “grainization” trend in the agricultural planting structure. Thus, the impact of fiscal decentralization on the area planted with grains varies with wage level, where higher wage levels can alleviate the negative effects of fiscal decentralization on grain planting.

Hypothesis 2:

Fiscal decentralization has a threshold effect on planting structure, and the “non-grainization” effect will weaken when the wage level is higher.

Under the fiscal system, local governments possess relatively independent fiscal and administrative powers, allowing them to adaptively manage financial resources based on local circumstances. However, changes in agricultural support structures can impact local government support for grain production. Specifically:

Level of agricultural fiscal support: This can reflect the extent of local government constraints by central financial support, particularly in major grain-producing but economically disadvantaged regions. With the abolition of agricultural taxes, local governments’ main incentive to focus on agriculture stems from central agricultural transfer payments, project-based transfer payments, and pressures from higher-level government assessments. Therefore, even with low tax benefits from agriculture, local governments remain motivated to invest in grain production to secure central transfer payments.

Impact of agricultural support funds: The level of financial support for agriculture also determines farmers’ enthusiasm for planting grains. In areas with higher levels of agricultural fiscal support, farmers’ willingness to cultivate grains increases, thus reducing the impact of fiscal decentralization on the “non-grainization” of agricultural structures. Therefore, with the central government increasingly emphasizing agriculture, variations in the level of agricultural fiscal support can influence the impact of fiscal decentralization on grain cultivation.

Hypothesis 3:

Fiscal decentralization has a threshold effect on planting structure, and the effect of “non-grainization” will be weakened when the intensity of agricultural support is higher.

Fiscal decentralization refers to the process by which the central government devolves certain fiscal powers to local governments, enabling them to better utilize their resources to foster local economic development. However, this decentralization may also lead to a “non-grain” effect, whereby local governments might prefer to allocate land resources to the cultivation of non-grain crops to achieve higher economic returns. Since the tax-sharing reform, local governments’ revenue has become heavily dependent on land-based fiscal revenues, which are bolstered by rapid urbanization and industrialization. Driven by urbanization and industrialization, surplus rural labor continuously migrates to cities, creating space for the large-scale operation of land transfers. Concurrently, land finance-driven urbanization and industrialization rapidly increase land prices, thereby boosting local governments’ enthusiasm for land consolidation [31]. The increase in land prices can potentially lead to an increase in farmers’ land income [14], which may reduce their dependency on grain production. This financial shift encourages farmers to allocate more land to non-agricultural sectors, such as industrial and service industries, diversifying their economic activities and potentially increasing their profitability. However, when land prices reach a certain level, the incentive for farmers to venture into non-agricultural industries may become constrained. High land costs can increase the risk associated with investments in non-agricultural sectors, particularly for small-scale farmers. On the other hand, as land prices soar, the government may intensify the regulation and control of land use to protect the land rights of farmers and prevent the adverse impacts of excessive development and external investment on the rural economy [32]. This regulatory oversight could potentially limit the growth of non-agricultural industries. In a framework of fiscal decentralization, local governments possess greater autonomy in determining land use policies, and their decision making is influenced by land prices [33,34]. Therefore, under the backdrop where local governments’ fiscal autonomy is dependent on land finance, changes in land finance can alter the impact of fiscal decentralization on grain cultivation.

Hypothesis 4:

Fiscal decentralization has a threshold effect on planting structure, and the “non-grainization” effect will be weakened when the land price is higher.

Finally, combined with related research content, the logical framework of this paper is shown in Figure 1.

3. Research Design

3.1. Model Construction

(1) Static panel model: To verify the hypotheses mentioned, this study constructs a static panel data model based on the theoretical background provided. The model is designed to assess the effects of fiscal decentralization on the structure of agricultural planting. The specification of the basic econometric model is as follows:

$$P{S}_{i,t}={\beta }_0+{\beta }_{1}F{\mathrm{D}}_{i,t}+{\beta }_2{X}_{i,t}+u_i$$

(1)

In the model, PS_i,t serves as the dependent variable, representing the agricultural planting structure of region i in year t, FD_i,t is the core explanatory variable, denoting the level of fiscal decentralization at the city level. The set of control variables, X, includes the index of technological advancement in grain production (TA), the efficiency index for grain production technology (TE), urban land expansion (CS), level of mechanization (MA), and labor migration (LA). u_i represents the composite error term.

(2) Dynamic panel model: The adjustments in agricultural planting structures led or guided by local governments are dependent on the variations in local resource endowments [35]. During the actual process of grain production, it is essential to consider several constraints: the costs associated with changing planting structures, the quality of the farmland, the available production technologies, and the willingness of farmers to engage in grain cultivation. The difficulty in adjusting these production factors contributes to the inherent ‘stickiness’ of grain production [36], where implementation of changes is slow and there is often resistance to rapid shifts. This stickiness in the agricultural sector necessitates a model that can account for these slow-moving dynamics effectively. Therefore, traditional estimation methods such as ordinary least squares (OLS), fixed effects (FE), and random effects (RE) can introduce biases due to their inability to fully account for dynamic changes and autocorrelation in data. To mitigate these issues, a dynamic model (model 2) is constructed on the foundation of model 1 by incorporating a lagged term of the proportion of grain planting area. This lagged variable helps control for “stickiness” in agricultural practices and addresses endogeneity concerns that may arise from the temporal dependencies of the variables.

The model is specified as follows, where δ represents fixed effects that are constant over time, and ε is the random error term:

$$P{S}_{i,t}={\beta }_0+{\lambda }_{1}P{S}_{i,t-1}+{\beta }_{1}F{D}_{i,t}{X}_{i,t}+{\beta }_2{X}_{i,t}+{\delta }_i+{\epsilon }_{i,t}$$

(2)

(3) To explore whether the “non-grain” effect of fiscal decentralization on agricultural planting structures is influenced by factors such as wage levels, land finance, and fiscal support for agriculture, and whether these influences exhibit threshold characteristics, a threshold effect model is constructed following the methodology proposed by Hansen [37]. Initially, it is essential to determine if a threshold characteristic exists; thus, the model is set up with the assumption of a single threshold characteristic. The model specification is as follows:

$$P{S}_{i,t}={\beta }_0+{\beta }_{1}F{D}_{i,t}I({\mathrm{g}}_{i,t}\le \gamma )+{\beta }_{2}Fi{n}_{i,t}I({\mathrm{g}}_{i,t}\ge \gamma )+{\beta }_3{X}_{i,t}+{\epsilon }_{i,t}$$

(3)

In the context of the dynamic panel model, g_i,t serves as the threshold variable, which may include wage levels, the intensity of land finance, or the extent of fiscal support for agriculture. The parameter γ represents the threshold value that is to be estimated. The function I (.) is an indicator function that evaluates to 1 when the expression inside the parentheses is true, otherwise, it evaluates to 0. This setup allows the model to adapt and include multiple thresholds, enhancing the analysis of how these varying conditions influence agricultural planting structures under fiscal decentralization.

3.2. Variable Description

(1) Core explanatory variable—fiscal decentralization (FD): Fiscal decentralization lacks a universally accepted measurement indicator, leading to significant empirical discrepancies due to variations in assessment methods. The literature classifies fiscal decentralization metrics into three broad categories: fiscal revenue decentralization, fiscal expenditure decentralization, and fiscal self-sufficiency. Fiscal self-sufficiency, in particular, reflects nuances in transfer payments and off-budget revenues and expenditures more effectively than the other metrics. Therefore, this paper adopts fiscal self-sufficiency as the measure of fiscal decentralization, defined as the ratio of local government revenue to local government expenditure.

(2) Dependent variable: This study focuses on the structure of grain planting, represented by the proportion of total sowing areas dedicated to wheat, rice, and corn. This variable highlights the allocation of agricultural resources towards the production of these staple grains.

(3) Threshold variable: The relationship between fiscal decentralization and agricultural planting structures may exhibit nonlinear effects due to regional variations. To analyze this, the study uses wage levels, fiscal support for agriculture, and land finance as threshold variables. Wage levels are expressed as the average wage of on-post workers at the city level, adjusted for price changes using the GDP deflator index (base year 2004) and then logarithmized. Fiscal support for agriculture is quantified by the proportion of expenditures on agriculture, forestry, and water affairs in total city-level government expenditures. Land finance is measured using the logarithm of per capita land concession fees at the city level.

(4) Control variables: ① Grain technical advancement index (TA), ② Grain technical efficiency index (TE), both of which are measured by the Malmquist index of DEA. On the one hand, the Malmquist index of DEA has been widely used in terms of technical efficiency after the theoretical basis of Charnes [34] and a non-parametric linear programming algorithm given by Fare [35]. On the other hand, the DEA–Malmquist index can be divided into the technical efficiency index (TE) and technological advancement index (TA) under the condition of constant return to scale. To apply the DEA index method, it is necessary to define the input and output variables first. The input variables in this study include the level of mechanization, chemical fertilizer input, effective irrigated area, crop sowing area, and the number of people employed in agriculture, forestry, animal husbandry, and fisheries. Following the calculation method used by Li Mingwen et al. [38], each input variable is multiplied by the ratio of grain sowing area to the total sowing area. The output variable is grain yield. This approach ultimately measures the technical efficiency index and technological advancement index in grain production. ③ Urban land expansion (ULE): This is measured by the ratio of the built-up area to the administrative area, which assesses the level of land urbanization. Following fiscal decentralization, as land finance becomes a core component of local revenue, regions are motivated to expand land use. This expansion often leads to a crowding-out effect that can significantly impact the support for agriculture and the area of arable land used for farming. ④ Level of mechanization (MA): The level of mechanization is quantified using the logarithm of the total power of agricultural machinery, which is a common measure. Since grain crops are more readily mechanized compared to cash crops, the ongoing urbanization and modernization of agriculture are expected to enhance the level of rural mechanization, thereby putting grain production at an advantage. Consequently, it is anticipated that higher mechanization levels will promote a structure in agricultural planting that favors grain production, termed “trend toward grain”. ⑤ Labor transfer (LA): Measured by the proportion of the urban population relative to the total municipal population. As labor continuously shifts from rural to urban settings, labor-intensive cash crops require higher human input, while grain crops (rice, corn, wheat) demand comparatively less labor and are more suited to mechanized farming. Thus, this shift is likely to facilitate the substitution of capital for labor, promoting an increase in the area dedicated to grain cultivation.

3.3. Data Sources

The indicators for this study predominantly derive from the EPS Database, “China City Statistical Yearbook”, “China Rural Statistical Yearbook”, and “China Land and Resources Statistical Yearbook”. Missing data were supplemented from municipal statistical yearbooks and statistical bulletins, and linear interpolation was employed for individual years missing data in some cities. Cities with extensive missing data were excluded from the analysis. Ultimately, a total of 170 prefecture-level cities from major grain-producing areas were selected as subjects for this study (based on the criteria outlined in the National Medium and Long-term Plan for Food Security (2008—2020) thirteen provinces—Liaoning, Jilin, Heilongjiang, Inner Mongolia, Hebei, Shandong, Anhui, Jiangsu, Jiangxi, Henan, Hunan, Sichuan, and Hubei—were designated as major grain-producing areas, excluding provincial municipalities directly under the central government and minority autonomous regions), encompassing five variables. To illustrate the scientific rigor in the selection of samples, data from four specific years—2006, 2011, 2016, and 2021—were chosen. The core explanatory and dependent variables were visualized using ArcGIS 10.8, as shown in Figure 2 and Figure 3.

The variance inflation factors (VIFs) for the variables were analyzed using the Stata 16 software. The results revealed that all variables had VIF coefficients less than 3, indicating no severe multicollinearity issues among the observed variables. This confirmation allows for the inclusion of the sample data in the regression model for further analysis. Descriptive statistics for the variables are presented in

Table 1. Descriptive statistics for variables.

Type	Variable	Obs	Mean	Std. Dev.	Min	Max	VIF
Core Explanatory Variable	FD	2380	0.476	0.215	0.055	1.541	2.22
Dependent Variable	PS	2380	0.71	0.13	0.10	0.99	-
Threshold Variables	SA	2380	6.920	0.096	6.464	8.196	2.83
	FA	2380	0.096	0.049	0.004	0.383	1.51
	LP	2380	6.635	1.278	0.952	10.187	2.93
Control Variables	TA	2380	1.056	0.357	0.246	2.307	2.73
	TE	2380	1.035	0.279	0.079	2.004	2.74
	ULE	2380	0.097	0.102	0.102	0.102	1.57
	MA	2380	5.561	0.878	1.768	7.621	1.02
	LA	2380	0.314	0.177	0.038	0.947	1.28

.

4. Empirical Results and Analysis

4.1. Analysis of the Baseline Model

Table 2. Regression results for the benchmark model.

Variable	Model (1) Static OLS	Model (2) Static RE	Model (3) Static FE	Model (4) System GMM
L.PS				0.746 *** (0.017)
FD	−0.187 *** (0.013)	−0.064 *** (0.015)	−0.052 *** (0.016)	−0.237 *** (0.027)
TE	−0.012 (0.012)	−0.023 *** (0.005)	−0.024 *** (0.005)	−0.072 *** (0.018)
TA	−0.042 * (0.023)	−0.064 *** (0.007)	−0.064 *** (0.008)	−0.212 *** (0.048)
ULE	0.058 *** (0.022)	−0.039 (0.023)	−0.044 * (0.025)	−0.087 * (0.043)
MA	0.017 *** (0.003)	0.009 (0.006)	0.010 (0.007)	0.008 * (0.006)
LA	0.188 *** (0.021)	0.055 ** (0.026)	0.038 (0.030)	0.109 *** (0.025)
Constant term	0.668 *** (0.048)	0.743 *** (0.039)	0.736 *** (0.042)	0.316 *** (0.057)
Sample size	2380	2380	2380	2038
AR (1)				0.000
AR (2)				0.817
R2/Hansen_overid	0.103	0.165	0.165	0.906

Note: (1) ***, **, * indicate that the statistical value is significant at the significance level of 1%, 5%, and 10%; (2) The R2 values of RE and FE are the R2 values estimated between groups.

presents the regression results for the baseline model. To effectively address potential endogeneity issues inherent in static panel estimations and to account for the ‘stickiness’ in agricultural planting structures, model 4 incorporates a first-order lag of the dependent variable. The System GMM approach proposed by Blundell et al. [37] was employed for estimation, accompanied by AR tests to assess the validity of the instrumental variables and Hansen tests to check for instrument overidentification. Consequently, the study utilizes the unofficial ‘xtabond2′ command for the System GMM estimation, using the first-order lag of the dependent variable as an instrumental variable, and reports the results of the AR and Hansen tests.

Table 2 displays the regression results for the baseline model. To validate the System GMM outcomes, the table also reports the results from static OLS, RE, and FE regressions that control for time and individual effects. Model (1) suggests that higher levels of fiscal decentralization at the city level are associated with a shift towards “non-grainization” agricultural planting structures. Given the geographical and climatic variations, individual city differences are significant; hence, the study further explores a static FE model under a dual fixed-effects framework. The results indicate that fiscal decentralization negatively impacts the “grainization” of agricultural planting structures, with an even higher coefficient magnitude. However, due to endogeneity issues related to the inclusion of lagged variables, a System GMM is applied to estimate the model, with the findings presented in

Table 3. Regression results for threshold effect.

Variable		F-Value	p-Value	Threshold Estimates	Lower Bound of Confidence Interval	Upper Limit of Confidence Interval
SA	Single threshold	65.04 **	0.013	9.302	9.285	9.320
	Double threshold	17.02	0.147	14.356	14.308	14.364
FA	Single threshold	13.36	0.450	0.018	0.011	0.019
LP	Single threshold	33.16 **	0.047	4.057	3.957	4.147
	Double threshold	9.30	0.477	7.028	6.523	7.037

Note: The critical values of 1%, 5%, and 10% in the table are the results obtained by “self-sampling” 300 times; ** is significant at 5% level.

′s model (4).

Firstly, the reported AR (1) is significant while AR (2) is not, suggesting no autocorrelation between the instrumental variables and the error terms, and the insignificant Hansen statistic confirms the validity and effectiveness of the instruments. Secondly, the parameter estimates indicate a significant negative impact of fiscal decentralization on the “grainization” of agricultural structures at a 5% significance level, implying that fiscal decentralization promotes a shift towards ‘non-grain’ crops. This outcome is robust. Intuitively, the economic performance-based assessment systems established by fiscal decentralization lead local governments to overlook agricultural production, reducing the enthusiasm for grain cultivation.

Examination of the control variables reveals that both grain production technical efficiency and technological advancement negatively impact the area dedicated to grain sowing. The reasons are twofold: on one hand, under the backdrop of labor migration, the mismatch between grain production technology levels and human capital leads to a reduction in the proportion of grain-sown areas. According to the appropriate technology theory, the technological choices of an economy are constrained by the factor endowment structure available during a given period. When human capital is insufficient to match the level of grain production technology, the efficiency of grain production is significantly compromised, thus affecting the sown area. On the other hand, improvements in technological levels and efficiency help local governments meet grain production targets, potentially leading to a decrease in their enthusiasm for intensive grain cultivation practices, such as double cropping of rice and intercropping with other grain crops, thereby affecting the grain sowing area. Urban land expansion also impacts the trend towards “grainization” of agricultural structures, though these findings are not robust. The levels of mechanization and labor migration both positively influence the “grainization” of agricultural structures. This is because cash crops, compared to grain crops, are more labor-intensive and require more human capital support. However, with the progression of urbanization and the development of the productive service market, the decreasing surplus rural labor force and increasing levels of mechanization are likely to promote the “grainization” of agricultural planting structures.

4.2. Analysis Based on Threshold Regression Models

Based on prior theoretical analysis and model estimations, this study further selects wage levels, land prices, and fiscal support for agriculture as threshold variables. Using Stata, a threshold regression analysis was conducted to explore the impact of fiscal decentralization on agricultural planting structures under varying conditions of wage levels, land prices, and fiscal support. According to the results presented in Table 3, both wage levels and land prices passed the single-threshold test, with threshold values identified at 9.302 and 4.057, respectively. However, fiscal support for agriculture did not pass the single-threshold test.

Table 4. Regression results show the impact of the threshold variable on the effect.

Variables	(5)	(6)
FD (SA ≤ γ)	−0.241 *** (0.027)
FD (SA > γ)	−0.071 *** (0.017)
FD (LP ≤ γ)		−0.225 *** (0.032)
FD (LP > γ)		−0.066 *** (0.016)

*** is significant at 1% level.

reveals the differential impacts of fiscal decentralization on agricultural planting structures at various wage levels. Specifically, when wage levels are below the threshold of 9.302, the effect of fiscal decentralization on the agricultural structure is negative (−0.241) and statistically significant at the 1% level. Conversely, when wage levels exceed this threshold, the negative impact of fiscal decentralization persists but the magnitude of the coefficient decreases, indicating that higher wage levels can mitigate the ‘non-grain’ effects of fiscal decentralization on agricultural structures. Similarly, when the fiscal measure related to land is below the threshold of 4.057, the impact of fiscal decentralization on agricultural planting structure is −0.225, significant at the 1% level. Above this threshold, the negative impact remains but the coefficient significantly decreases, suggesting that growth in land-related fiscal measures also helps alleviate the negative effects of fiscal decentralization on agricultural structure.

4.3. Sub-Sample Threshold Regression Model Analysis

Given the vast expanse of China and the diversity in resource endowment across different regions, especially in terms of climate and terrain for crop cultivation, there are notable differences in the planting of wheat, rice, and corn in the main grain-producing areas. To further explore whether the areas planted with different crops are influenced by fiscal decentralization, the study categorizes the main grain-producing areas into primary regions for wheat, corn, and rice. Additionally, wage levels, land prices, and fiscal support for agriculture are selected as threshold variables to examine the impact of fiscal decentralization on the cultivation of wheat, corn, and rice. (The main wheat producing areas were Anhui, Hebei, Henan, Hubei, Jiangsu, and Shandong provinces. The main corn producing areas were selected in Anhui, Heilongjiang, Hubei, Hunan, Jiangsu, Jiangxi, and Sichuan. The main rice producing areas were Hebei, Henan, Heilongjiang, Jilin, Liaoning, Inner Mongolia Autonomous Region, Shandong, and Sichuan.)

From the results shown in

Table 5. Results of the sub-sample threshold effect test.

Variable			F-Value	p-Value	Threshold Estimates	Lower Bound of Confidence Interval	Upper Limit of Confidence Interval
Wheat	SA	Single threshold	104.73 ***	0.000	13.801	11.938	13.842
		Double threshold	69.62 ***	0.000	14.934	14.921	14.938
	FA	Single threshold	117.97 ***	0.000	0.016	0.015	0.017
		Double threshold	11.16	0.373	0.069	0.067	0.070
	LP	Single threshold	46.20 **	0.001	6.509	6.482	6.521
		Double threshold	35.18 **	0.027	8.395	8.347	8.403
Corn	SA	Single threshold	82.13 ***	0.000	14.386	14.359	14.393
		Double threshold	25.75 *	0.063	14.982	14.968	14.991
	FA	Single threshold	56.93 ***	0.000	0.074	0.071	0.074
		Double threshold	41.71 **	0.043	0.188	0.181	0.189
	LP	Single threshold	78.00 ***	0.000	4.974	4.948	5.087
		Double threshold	19.12	0.187	7.066	6.990	7.081
rice	SA	Single threshold	5.60	0.663	14.396	14.391	14.398
	FA	Single threshold	15.76	0.157	0.101	0.098	0.10
	LP	Single threshold	13.23	0.210	6.458	6.154	6.469

***, **, * are significant at 1%, 5%, and 10% levels.

, in the wheat cultivation sector, wage levels, fiscal support for agriculture, and land finance all demonstrate threshold effects, with wage levels and land finance exhibiting dual threshold effects. In the corn sector, wage levels, fiscal support for agriculture, and land prices also display threshold effects, with wage levels and land prices showing single threshold effects, and fiscal support structures demonstrating dual threshold effects. In rice cultivation, significant threshold effects are not observed, and the coefficients are not significant; therefore, the impact of fiscal decentralization on rice cultivation is not listed in this document.

In wheat cultivation, threshold effects were observed in the impacts of wage levels, fiscal support for agriculture, and land finance on the “non-grainization” effect of fiscal decentralization, as seen in

Table 6. Regression results: Impact of sub-sample threshold variable.

	Wheat				Corn
Threshold Value	(7)	(8)	(9)	Threshold Value	(10)	(11)	(12)
FD (SA ≤ γ1)	−0.098 *** (0.018)			FD (SA ≤ γ1)	−0.101 *** (0.020)
FD (γ1 < SA ≤ γ2)	−0.039 *** (0.013)			FD (γ1 < SA ≤ γ2)	−0.071 ** (0.019)
FD (γ2 < SA)	−0.012 (0.013)			FD (γ2 < SA)	−0.028 (0.019)
FD (FA ≤ γ)		0.023 (0.110)		FD (FA ≤ γ1)		−0.065 *** (0.019)
FD (γ < FA)		−0.044 *** (0.015)		FD (γ1 < FA ≤ γ2)		−0.022 (0.019)
				FD (γ2 < FA)		0.179 *** (0.034)
FD (LP ≤ γ1)			−0.055 *** (0.016)	FD (LP ≤ γ)			−0.168 *** (0.022)
FD (γ1 < LP ≤ γ2)			−0.029 ** (0.016)	FD (γ < LP)			−0.068 *** (0.020)
FD (γ2 < LP)			−0.001 (0.015)

***, ** are significant at 1%, 5% levels.

. With respect to wage levels, as regional wages increase, the impact of fiscal decentralization on the “non-grainization” of agricultural planting structures gradually diminishes. When the wage level crosses the first threshold of 13.801, the negative impact of fiscal decentralization decreases from −0.098 to −0.039, and upon surpassing the second threshold of 14.934, the effect becomes statistically insignificant. This confirms that as regional wage levels rise, the influence of fiscal decentralization on the “non-grainization” of agricultural planting structures is reduced.

From the perspective of fiscal support for agriculture, when the structure of municipal support for agriculture is below 0.016, the impact of fiscal decentralization on wheat planting structures is not significant. However, as fiscal decentralization exceeds the first threshold, the proportion of wheat sown area is negatively affected. A possible explanation is that subsidies for wheat cannot compensate for the increases in labor and land costs, thereby reducing the area sown with wheat. According to the ‘National Compilation of Costs and Returns of Agricultural Products’, the cost per mu of wheat planting rose from RMB 355.92 in 2004 to RMB 1007.64 in 2017, an increase of 1.87 times, with labor costs rising from RMB 111.84 to RMB 361.87, and land costs from RMB 43.8 to RMB 207.1. Meanwhile, the net profit per mu of wheat decreased from RMB 169.58 in 2004 to RMB 0.7 in 2017, with increasing volatility. Thus, even as fiscal support for agriculture intensifies, the comparative disadvantage of wheat cultivation leads to further crowding out of wheat sowing areas due to the growth in fiscal decentralization.

Regarding land finance, as it increases, the negative impact of fiscal decentralization on wheat planting areas continuously decreases. Once land finance exceeds the second threshold of 8.395, the negative impact of fiscal decentralization on wheat planting areas becomes insignificant.

In corn cultivation, threshold effects are evident in the relationships between wage levels, agricultural fiscal support structures, and land prices with the “non-grainization” impact of fiscal decentralization, as detailed in Table 6. Regarding wage levels, a single threshold effect is observed in the impact of fiscal decentralization on agricultural planting structures. When wage levels exceed the threshold of 14.386, the influence of fiscal decentralization on the “non-grainization” of agricultural structures significantly decreases. In terms of fiscal support for agriculture, the impact of fiscal decentralization on corn cultivation becomes insignificant when the proportion of fiscal support is between 0.074 and 0.188. Beyond this range, as the ratio exceeds 0.188, fiscal decentralization begins to positively influence the area dedicated to corn cultivation. Regarding land finance, once the level surpasses the threshold of 4.974, the negative impact of fiscal decentralization on the area sown with corn significantly diminishes.

4.4. Robustness Test

To further confirm the robustness of the impact of fiscal decentralization on agricultural planting structures, robustness checks of the econometric models are necessary. Consequently, this section employs an alternative explanatory variable method for robustness testing. The core explanatory variable in this study is the level of government fiscal decentralization. Among the commonly used fiscal decentralization metrics—fiscal revenue decentralization, fiscal expenditure decentralization, and fiscal self-sufficiency—the latter, influenced predominantly by transfer payments, may not fully reflect the autonomy of local governments, potentially leading to biased results [39]. Thus, fiscal revenue decentralization is chosen as the substitute variable and, considering regional population differences, it is calculated per capita at the city level using the following Formula (4). Additionally, some literature reviews have critiqued the adequacy of fiscal decentralization indicators and considered the impact of the exogenous policy “provincial administration of counties” fiscal system reform. Innovatively, this study uses the proportion of provincially administered counties to depict the fiscal system at the city level under provincial administration [40]. The specific formula is shown as Equation (5). During the sample period, the content of provincial administration of county reforms mainly involved the devolution of fiscal authority to counties, including transfer payments, budget allocation, and year-end settlements, directly facilitating fiscal connections between provinces and counties. This reform granted local governments greater fiscal and economic autonomy, significantly enhancing their incentives for economic development. Consequently, an increase in the proportion of provincially administered counties amplifies the overall level of fiscal decentralization within city jurisdictions, thereby impacting the agricultural planting structure.

Municipal Fiscal Income Decentralization (FDI) = Per capita municipal fiscal income/(Per capita municipal fiscal income + Per capita provincial fiscal income + Per capita central fiscal income)

(4)

Proportion of Provincially Administered Counties (The statistics of the number of counties directly administered by the Ministry of Finance, according to the provincial documents listed in the Wikipedia entry “Provincial directly administered counties”, take June as the line of demarcation, take June as the line of demarcation, if the implementation of “provincial directly administered counties” in June or before June, then calculate from the current year, if he took office after June, then calculate from the following year) = Number of municipally administered counties under provincial governance/Total number of counties and county-level cities in the municipality

(5)

As shown in

Table 7. Robustness tests: Transforming a measure of municipal fiscal decentralization.

Variable	(13) Static OLS	(14) Static RE	(15) Static FE	(16) GMM	(17) Static OLS	(18) Static RE	(19) Static FE	(20) System GMM
L.PS				0.815 *** (0.026)				0.905 *** (0.037)
FDI	−0271 *** (0.023)	−0.164 *** (0.030)	−0.158 *** (0.033)	−0.263 *** (0.050)
Provincial administration of counties reforms					−0.058 *** (0.006)	−0.030 *** (0.009)	−0.028 *** (0.006)	−0.012 * (0.013)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant term	0.675 *** (0.042)	0.751 *** (0.039)	0.748 *** (0.042)	0.397 *** (0.083)	0.694 *** (0.046)	0.735 *** (0.063)	0.730 *** (0.041)	0.274 *** (0.063)
Sample size	2380	2380	2380	2038	2380	2380	2380	2038
AR (1)				0.000				0.000
R2/AR (2)				0.617				0.725
Hansen_overid	0.083	0.170	0.170	0.986	0.074	0.170	0.170	0.830

***, * are significant at 1% and 10% levels.

, the impacts of fiscal revenue decentralization and the “provincial administration of counties” reforms on the agricultural planting structure are negative and significant. Additionally, the results from the System GMM are also negative and significant. Both sets of findings are consistent with the baseline empirical model, enhancing the persuasive power of the baseline empirical evidence.

5. Discussion

Food security is key to alleviating poverty and responding to the shocks of uncertainty. Many studies have shown that farmers have different opinions about whether to plant food. The basic consensus is that food is the agricultural product with the highest public value and the lowest comparative return. As a result, the main debate has focused on who should support farmers and what role the central and local governments should play. This study measures the fiscal decentralization relationship between central government and local government by introducing the fiscal self-sufficiency rate. The study confirms that the higher the degree of fiscal decentralization, the more serious the “non-grainization” of the planting structure. This coincides with the studies of Chari et al. [41] and Gong, B. [42]. The above conclusions show that agricultural production, especially food production, has a limited role in economic growth, and agricultural development has been weak in the evaluation and supervision system of local officials for a long time, and some local governments lack incentives to develop food production. In the context of power expansion, financial resources show a trend of leaving agriculture, infrastructure construction is insufficient, modern grain production factors are replaced slowly, and total factor productivity is relatively declining, forming a “push” for grain production [43].

Further, in the process of discussing the mechanism of fiscal decentralization on grain planting structure, this paper found that fiscal decentralization may have nonlinear effects. First of all, the increasing proportion of local finance will lead to local economic development and industrial upgrading, which is conducive to non-agricultural employment of local farmers. From the perspective of farmers’ part-time employment, the planting link of food crops is relatively simple, and the optimal decision in the process of farmers’ non-farm employment is to cultivate food simultaneously. Therefore, in the process of increasing the level of non-farm labor, fiscal decentralization has a threshold effect on the “non-grainization” planting structure. Secondly, the increase in local financial support for agriculture will also encourage farmers to plant food, thus triggering a threshold effect between the two [44]. Moreover, the higher the degree of fiscal decentralization, the local government may increase the supervision and regulation of land use, so as to protect farmers’ land rights and interests and prevent social stability problems that may arise from over-development and external investment [45]. Such regulation could limit the growth of non-grainization industries. The above different influence paths confirmed the nonlinear relationship between fiscal decentralization and grain planting structure, and expounded the idea of alleviating “non-grainization” through different empirical perspectives.

To sum up, optimizing the planting structure in major grain-producing areas is crucial not only for ensuring food security strategy but also for advancing modern agricultural development. Beyond exploring more scientific methods of increasing yield, there is a pressing need to accelerate the optimization of policy and regulatory tools to further strengthen and enhance the level of food security in these regions. Based on the findings of this study, the following policy implications are proposed. (1) Moderating fiscal decentralization to stabilize grain planting proportions: It may be beneficial to slightly reduce the extent of fiscal decentralization to stabilize the area dedicated to grain cultivation. The central government should emphasize an approach that aligns responsibilities with funding (“funding by responsibilities” and “responsibilities by funding”) to clarify the distribution of powers among various levels of government involved in grain production. This strategy aims to fundamentally ensure the maximum efficacy of fiscal resources and achieve standardized regulation of grain production across regions. (2) Addressing threshold effects of fiscal decentralization on agricultural structures: It is crucial to be aware of the potential threshold effects fiscal decentralization may have on the structure of grain cultivation. Appropriately increasing regional wage levels and land finance can mitigate the impacts of fiscal decentralization that drive the shift away from grain cultivation. However, it is essential to recognize that the mitigating effect of land finance often stems from the encroachment on agricultural land, a practice that is inherently unsustainable. Therefore, further rational constraints and regulations are necessary to moderate local government actions regarding land finance. It is also vital to prevent an excessive dependency on land finance in major grain-producing areas to ensure that it positively contributes to stabilizing grain production. (3) Enhancing the role of wage levels, land finance, and fiscal support in stabilizing wheat and rice cultivation: From the perspective of the internal structure of grain cultivation, it is vital to maximize the roles of wage levels, land finance, and fiscal support to stabilize the cultivation of wheat and rice, particularly for wheat. On one hand, the provision of wheat seed subsidies should be implemented to reduce the cost burden on farmers purchasing seeds. On the other hand, subsidies for agricultural inputs such as fertilizers, pesticides, and agricultural machinery should be provided to help lower production costs and enhance crop quality and yield. Additionally, the government could introduce further policies such as converting grain to fodder, price subsidies, and purchasing protections to ensure farmers receive fair and reasonable price returns. These measures would enhance farmers’ enthusiasm for cultivating wheat, stabilize the basic income from wheat cultivation, and mitigate the negative impact of fiscal decentralization on the proportion of wheat planted, thereby promoting sustainable agricultural development.

It is worth mentioning that due to the availability of data in this paper, we only analyzed the relationship between fiscal decentralization and grain planting structure within 170 major grain-producing areas from 2004 to 2017, and did not group control areas outside the major grain-producing areas, which means the conclusions of this paper may be affected by certain self-selection. In the future, we will conduct further analysis of samples over longer time series in different regions.

6. Conclusions

This study employed static models and System GMM models to perform empirical tests using panel data from 170 prefecture-level cities in China’s main grain-producing regions from 2004 to 2017. This analysis quantitatively explored the impact of fiscal decentralization on the structure of agricultural planting. The primary conclusions of this research are as follows:

(1) Fiscal decentralization has significantly reduced the proportion of grain cultivation in China’s major grain-producing regions. Under fiscal decentralization, local governments possess greater fiscal and administrative authority, which may lead them to prioritize the development of the local economy and the enhancement of fiscal revenues. In this context, local governments might prefer to allocate resources to more economically profitable sectors rather than grain crop production. Such decisions could lead to a reduction in the area allocated to grain cultivation, thereby negatively impacting food security. Moreover, local governments with substantial fiscal power have greater autonomy in formulating agricultural policies. If local authorities neglect food security needs or overly focus on economic benefits, these policies could adversely affect the structure of grain planting.

(2) Wage levels and land finance both exhibit threshold effects, with the “non-grainization” impact of fiscal decentralization on agricultural planting structures gradually weakening as both increase. Specifically, the threshold effect of wage levels suggests that as wages rise, farmers’ investment in and enthusiasm for grain cultivation may gradually diminish. Higher wages imply increased income opportunities in other industries, leading farmers to potentially choose employment in other sectors for better earnings. This trend could enhance the “non-grainization” effect in agricultural planting structures.

Furthermore, the threshold effect of land finance indicates that as land finance increases, government support and investment in agriculture may progressively decrease. An increase in land finance means that more government funds are allocated to urban construction and industrial development, potentially squeezing out agricultural investments. This trend could lead to a decrease in farmers’ investment in and enthusiasm for grain cultivation, further exacerbating the “non-grainization” effect of agricultural planting structures. Additionally, as wage levels and land finance rise, the “non-grainization” effect of fiscal decentralization on agricultural structures progressively weakens. This indicates that with economic development and societal progress, government support and investment in agriculture are also gradually increasing. Simultaneously, as wage levels rise, farmers’ investment in and enthusiasm for grain cultivation also gradually improve. These factors collectively act to weaken the “non-grainization” effect of fiscal decentralization on agricultural planting structures.

(3) Regarding the internal structure of grain cultivation, fiscal decentralization negatively impacts both wheat and corn cultivation, with observable threshold effects linked to wage levels, fiscal support for agriculture, and land finance. This outcome suggests that fiscal decentralization leads to reduced governmental investment in agriculture at the local level, thereby affecting the cultivated area and yield of wheat and corn. Furthermore, fiscal decentralization may cause deviations in the formulation and implementation of agricultural policies, exacerbating the negative impact on wheat and corn cultivation.

Additionally, the existence of threshold effects associated with wage levels, fiscal support for agriculture, and land finance indicates that these factors can conditionally trigger threshold effects on the impact of fiscal decentralization on wheat and corn cultivation. Specifically, as wage levels increase, farmers may prefer to work in other industries for higher incomes, reducing their investment in and commitment to grain cultivation. Changes in fiscal support for agriculture and land finance also influence farmers’ willingness and ability to invest in grain cultivation.

(4) The increase in the fiscal support structure for agriculture enhances the negative impact of fiscal decentralization on wheat cultivation, while the effect is opposite for corn cultivation. This suggests that under fiscal support policies, corn exhibits higher adaptability compared to wheat. Specifically, fiscal decentralization might lead local governments to focus more on short-term benefits, thereby favoring the cultivation of certain crops. Corn is more drought-resistant and can thrive in poorer soils, making it viable for a broader range of regions. Moreover, the market demand for corn is relatively high, especially in the livestock and bioenergy sectors. Consequently, local governments may prefer to support corn cultivation to boost local economic development.